Documentation

Tenant OwnerReviewerOperator

Product knowledge base, playbooks, and embedded guidance for every persona.

AI Reviewer

CoffeeBreak includes an AI-powered code reviewer that helps your team review pull requests more quickly and consistently.
The AI Reviewer works alongside your human reviewers to improve quality and reduce turnaround time.


๐Ÿค– What the AI Reviewer Does

When CoffeeBreak opens or updates a pull request, the AI Reviewer:

  • Examines the changes in the PR (diff + surrounding context)
  • Highlights potential issues
  • Suggests improvements
  • Flags risky or unclear sections of code
  • Adds comments directly to the PR
  • Provides a summary of the changes

It functions like an additional reviewer on your team.


โ— What the AI Reviewer Does Not Do

To set expectations clearly:

  • It does not auto-merge pull requests
  • It does not override human reviewers
  • It does not enforce blocking checks unless your workspace enables them
  • It does not run tests, CI, or build tools
  • It does not analyze your entire repository (only the diff + limited context)

Human reviewers always have final authority.


๐Ÿง  How the AI Reviewer Helps

You can expect benefits such as:

โœ“ Faster review cycles

It catches issues early, reducing back-and-forth.

โœ“ More consistent quality

It applies the same review standards across all tasks and contributors.

โœ“ Better understanding of generated changes

It explains what the agent changed and why.

โœ“ Reduced reviewer load

Your team can focus on high-level decisions instead of mechanical issues.


๐Ÿ” When the AI Reviewer Runs

The reviewer is automatically invoked when:

  • CoffeeBreak opens a new pull request
  • The system updates a PR after refinement
  • A task transitions into a review-required state
  • A human reviewer requests changes and CoffeeBreak proposes a revision

You do not need to trigger it manually.


๐Ÿงฉ Using AI Review Comments

AI Reviewer comments appear just like human comments:

  • Inline notes on specific lines
  • General summary comments
  • Suggestions for improvements
  • Alerts on missing tests or potential bugs

You can:

  • Reply
  • Resolve
  • Request changes
  • Use them to guide your own review

โš™๏ธ Workspace Configuration

Your workspace may optionally configure:

โœ“ Confidence thresholds

Set when AI Reviewer findings require human approval.

โœ“ Required reviewers

Specify when human review is mandatory regardless of AI confidence.

โœ“ Auto-merge rules

The AI Reviewer never merges PRs on its own.
Workspace rules always apply.


๐Ÿ” Example AI Reviewer Comments

Here are typical comments you might see:

  • โ€œConsider renaming this method for clarity.โ€
  • โ€œPossible null reference โ€” should this be validated?โ€
  • โ€œLogic in this block is complex; consider simplifying.โ€
  • โ€œTests are missing for this new behavior.โ€
  • โ€œThis loop looks identical to code in X; consider extracting a helper.โ€

๐Ÿ‘ค Human Review Still Matters

The AI Reviewer is meant to assist, not replace, your reviewers.

CoffeeBreak always respects:

  • Branch protection rules
  • Code owner requirements
  • Approval thresholds
  • Reviewer groups
  • Any restrictions your organization has in place

๐Ÿ”’ Privacy & Security

All reviews stay within your workspace and are never shared across organizations.
The reviewer only sees the code in your repository and the changes generated for your tasks.


๐Ÿงช For Marketplace Developers

If you are building custom agents that generate code:

  • The AI Reviewer will automatically analyze their output
  • Higher-quality outputs lead to better routing and refinement
  • No integration work is required on your part

See also:

  • external/CreateACustomAgent.md
  • external/AgentMetadataSpec.md